In voice and data communications networks, there is an on-going need to minimize bandwidth requirements and improve the quality of voice or data traffic. Reducing the bandwidth is typically achieved by implementing compression algorithms to remove redundancy from a signal. On the other hand, quality is typically improved by adding redundancy to a signal by, for example, implementing error detection and correction techniques, and by recovering from errors by using lost frame concealment techniques.
Conventional systems attempt to achieve a balance between bandwidth and quality by using a combination of methods. In a conventional system, at the transmitting side, a source coder/quantizer is provided to quantize and compress the signal to be transmitted, i.e. reduce the bandwidth required, while a channel coder is provided to add information for use in error detection and correction, i.e. improve quality. The signal then travels through a channel (data link) where it may be corrupted. At the receiving side, a corresponding channel decoder and source decoder are provided to decode the signal received.
One of the issues in communication systems is that, as the interference level increases, the quality of recovered signal falls off rapidly. One conventional approach to overcome this problem has been the use of adaptive source/channel coding (e.g. GSM's Adaptive Multi-Rate [AMR]). Adaptive source/channel coding allows a variation in the level of source coding based on the amount of interference found on the channel data link. For example, a lower level of source coding is performed when the level of interference is high. This allows for more redundancy in the signal and thus, the interference will have less impact on the signal. However, this also has the effect of increasing bandwidth requirements. In a similar way, when the level of interference is low, a higher level of source coding can be used. In this way, adjustments can be made adaptively to counteract the effects of interference during signal transmission.
While adaptive source/channel coding adjusts the source coder based on interference conditions, other conventional approaches are directed to the receiver side of the channel. In a communication system, when a data bit is received, there is some uncertainty as to whether or not the bit is a 1 or a 0 due to distortion, interference, noise on the channel, or the like. In a conventional system, the channel decoder would typically examine an incoming signal and make a hard decision, that is, a final decision as to whether a particular bit received is a 1 or 0 without reference to any information other than the received bit. Thus, the output of the channel decoder would only be 0 or 1 (the output is quantized to two levels), which are called “hard-bits”.
More recently, the concept of using “soft bits” has been introduced. A soft bit represents a probability that a particular bit will be either a 1 or 0. The concept of soft bits also often involves the use of various types of available information (sometimes called a-priori information) to assist with interpretation of a data stream as opposed to merely examining a particular data bit or codeword that is being processed. As a very simple example, if we assume that at the sending end, a signal value of 1V represents a bit value of 1 and −1V represents a bit value of 0, when making a hard decision for an incoming bit, a receive signal value of 0.75 Volts may be assigned a bit value of 1 and −0.25 Volts may be assigned a bit value of 0. However, in such a scenario it is difficult to assign a bit value to a signal value of 0.0 Volts. Such a determination can be more effectively made by considering a-priori information, for example, if the type of signal is known to have a higher percentage of 1's than 0's, then there may be a greater probability that a received signal of 0.0 Volts should be classified as a 1 and not a 0. As such a received bit can be quantized to a larger number of levels. Other types of information used in determining soft bit probabilities may relate to previous bits received, or, with a delay, following bits received. Overall, this additional information can be used to produce a more accurate estimate of the transmitted bit value.
A channel decoder may incorporate soft bits to become a soft channel decoder, which outputs a soft bit (i.e. a probability) rather than a hard bit (i.e. a 1 or 0). In particular, the soft channel decoder may include forward error checking (FEC) decoding with a soft-in, soft-out maximum a-posteriori probability (MAP) decoder.
In this type of system, the soft channel decoder adjusts the probabilities based on a-priori knowledge to compute adjusted output probabilities. These adjusted probabilities are then provided to a soft source decoder, which uses the probabilities to determine codeword probabilities. These codeword probabilities can be used to compute the source decoder output based on, for example, a most likely output using maximum a-posteriori probability (MAP) or by weighting the corresponding decoder outputs using a minimum mean square (MMS) approach.
Information regarding soft bits and the types and usage of a-priori information for both speech and images can be found in the following research papers and the documents that they reference. The following documents are hereby incorporated herein by reference:    T. Fringscheidt and P. Vary, Softbit Speech Decoding: A New Approach to Error Concealment, IEEE Transactions on Speech and Audio Processing, Vol. 9, No. 3, March, 2001.    F. Lahouti, A. Khandani, Soft Reconstruction of Speech in the Presence of Noise and Packet Loss, Technical Report UW-E&EC#2003-4, University of Waterloo, May 15, 2003.    A. Mertins, Image Recovery from Noisy Transmission using Soft Bits and Markov Random Field Models, University of Oldenburg School of Mathematics and Natural Sciences, GERMANY, http://www.uni-oldenburg.de/sigproc/papers/mertins-oe-oct03.pdf, October 2003.    J. Kliewer, N. Gortz, Soft-input Source Decoding For Robust Transmission Of Compressed Images Using Two-Dimensional Optimal Estimation, University of Kiel, Proceedings ICASSP 2001, vol. IV, pp. 2565-2568, Salt Lake City, Utah, USA, May 2001, http://www.Int.ei.tum.de/mitarbeiter/goertz/icassp2001.pdf.
The above conventional approaches to balancing bandwidth and quality requirements have various issues. For example, GSM's AMR solution requires complex control procedures to allow for adaptive control of the source coders and corresponding source decoders based on interference levels. Further, methods of error checking and lost frame concealment typically require increases in bandwidth requirements and processing capability.
As such, there is a need for an improved system and method for encoding and decoding information signals using soft decoding.